Server management apparatus and server management method

Abstract

A server management apparatus for lowering migration costs during scale-in / scale-out and workload consolidation of a cluster system(s) to thereby reduce power consumption is disclosed. The apparatus manages a physical server group which renders operative a virtual server group thereon and, when putting into practice a cluster system including a plurality of virtual servers placed in the physical server group, manages the layout state of virtual servers pursuant to the load state of the virtual server group. When executing scale-in, a virtual server operating on a physical server with the minimum number of operative virtual servers is specified as a shutdown target. When executing scale-out, a workload variation is predicted to control a scale-out destination of cluster system so that load variation-resembled cluster systems gather on the same physical server. The scale-in execution timing is delayed if the predicted load variation tends to rise and accelerated if it falls.

Description

INCORPORATION BY REFERENCE[0001]The present application claims priority from Japanese application JP2009-245035 filed on Oct. 26, 2009, the content of which is hereby incorporated by reference into this application.BACKGROUND OF THE INVENTION[0002]1. Field of the Invention[0003]The present invention relates to server virtualization technologies for allowing a single physical server to operate as one or more independent virtual servers. This invention also relates to a technique for dynamically modifying the configuration of a cluster system which is arranged by virtual servers to thereby reduce electrical power consumption.[0004]2. Description of the Related Art[0005]In recent years, aggressive attempts are made to promote the introduction of virtualization technology into business-entity information systems (referred to hereinafter as corporate systems), which are information systems for use in business enterprises or companies. By the virtualization technology, physical or “real” ...

AI Technical Summary

Benefits of technology

[0014]A first problem to be solved lies in the necessity to control a scale-in / scale-out target of the cluster system in such a way that the number of virtual server migration events is minimized during workload consolidation. In the system taught from US2005 / 0060590A1, an increase in virtual server migration number during the workload consolidation results in a decrease in power consumption reduction effect and also in occurrence of a decrease in response performance due to unwanted increases in virtual server migration costs.

[0028]According to this invention, it is possible to reduce the migration costs to the greatest possible extent at the time of cluster system's scale-in / scale-out or workload consolidation—i.e., the number of system configuration modifications—to thereby reduce power consumption.

Problems solved by technology

A first problem to be solved lies in the necessity to control a scale-in / scale-out target of the cluster system in such a way that the number of virtual server migration events is minimized during workload consolidation.

In the system taught from US2005 / 0060590A1, an increase in virtual server migration number during the workload consolidation results in a decrease in power consumption reduction effect and also in occurrence of a decrease in response performance due to unwanted increases in virtual server migration costs.

Note that the virtual server migration costs are a workload of central processing unit (CPU), input / output (I / O) load, migration time, etc. increases in CPU load and I / O load result in an increase in power consumption.

Additionally, an increase in migration time results in shortening of a time taken to power off a physical server, which leads to a decrease in power consumption reduction effect.

Further, the increase in migration time retards execution of a load variation tracking operation, thereby making it impossible for the cluster system to offer the response performance required by users.

On the other hand, Dara Kusic et al. is silent about the workload consolidation processing; so, it does not employ a method of modifying the configuration of a cluster system in such a manner as to minimize the virtual server migration number during the workload consolidation.

Due to this fact, mere combination of the techniques of US2005 / 0060590A1 and Dara Kusic et al. is encountered with the problem of a decrease in power consumption reduction effect due to virtual server migration costs and the problem of a decrease in response performance.

Images